Show HN: Freestyle – Wispr Flow Alternative that respects privacy
Freestyle is an open-source voice dictation app created as a privacy-focused alternative to Wispr Flow. It allows users to run transcription models like Whisper, Qwen, and Sensevoice locally on their device, ensuring that voice data never leaves the user's machine.
Background
- Wispr Flow is a commercial voice dictation tool (macOS/Windows) that transcribes speech to text in any app. It is cloud-only: audio is sent to external servers for processing, raising privacy concerns for users who don't want their speech data leaving their machine.
- Freestyle is an open-source alternative that runs transcription models locally on the user's own computer. It supports models like Whisper (OpenAI's open-source speech-to-text model), Qwen (Alibaba's large language model family, which includes speech capabilities), and SenseVoice (an open speech model from the Chinese AI lab SenseTime).
- The key trade-off: local transcription is slower and uses more of the user's own hardware (CPU/GPU) than cloud-based services, but it guarantees that no audio data ever leaves the device.
- This is part of a broader trend in 2024–25 where developers are building desktop tools that replace cloud-dependent AI services with fully local, privacy-preserving open-source alternatives.